1 / 53

Detector Simulation & GRID Technology

Detector Simulation & GRID Technology. F.Carminati Erice, September 28, 2003. Total weight 2,000t Overall length 17.3m. Total weight 53,000t Overall length 270.4m. Alice collaboration. online system multi-level trigger filter out background reduce data volume. Total weight 10,000t

emile
Download Presentation

Detector Simulation & GRID Technology

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Detector Simulation& GRID Technology F.CarminatiErice, September 28, 2003

  2. Total weight 2,000t Overall length 17.3m Total weight 53,000t Overall length 270.4m Alice collaboration online system multi-level trigger filter out background reduce data volume Total weight 10,000t Overall diameter 16.00m Overall length 25m Magnetic Field 0.4Tesla 8 kHz (160 GB/sec) level 0 - special hardware 200 Hz (4 GB/sec) level 1 - embedded processors 30 Hz (2.5 GB/sec) level 2 - PCs 30 Hz (1.25 GB/sec) data recording & offline analysis Innovative Detector for Supercolliders

  3. ALICE Event/100 Innovative Detector for Supercolliders

  4. The question of simulation • Simulation is vital for HEP to evaluate the performance of the detector and estimate background BUT • Using GEANT 3.21 • Stay with FORTRAN, old physics and geometry • Using GEANT 4 • Not yet completely validated • Using FLUKA • Good physics but limited user interface We are in trouble! Innovative Detector for Supercolliders

  5. Geant3.tar.gz includes an upgraded Geant3 with a C++ interface Geant4_mc.tar.gz includes the TVirtualMC <--> Geant4 interface classes G3 G3 transport User Code VMC G4 G4 transport FLUKA transport FLUKA Reconstruction Geometrical Modeller Visualisation Generators The Virtual MC Innovative Detector for Supercolliders

  6. 3 million volumes ALICE Innovative Detector for Supercolliders

  7. Issues in simulation • Reliability of the MC • The simulation result are (at best!) as reliable as the MC used! • The problem is how well a MC describe a non existing detector in an inexperienced domain of energy • Multi-million €choices depend on this • Availability of adequate computing power • Cannot beat the 1/√n law • Variance reduction techniques are not commonplace in all HEP MC’s, and they are not of general use Innovative Detector for Supercolliders

  8. Reliability • The question usually asked is what is the precision needed for the different processes • Unfortunately this question is ill posed • The interaction between processes is highly non linear • It is almost impossible to determine “a priori” which process is important for a given case • The best strategy is • Describe processes “as well as possible” (sic!) • Constantly control the quality of the MC with experimental data • BTW, the same applies with cuts Innovative Detector for Supercolliders

  9. Boundary Single scattering through the boundary Affects e/π in finely segmented media Do we know everything about multiple scattering? No initial deviation? Innovative Detector for Supercolliders

  10. Parametrisation and physics • All MCs contain parametrisations to be tuned… however • If these are the parameters of a spline, no application to a set of data different from the original is allowed • This is the case of GHEISHA (page 3 of the GHEISHA manual), still largely used for LHC simulations! • If these are parameters of a physic model, a little more optimism is allowed • This is the case for FLUKA and the latest developments in GEANT4 Innovative Detector for Supercolliders

  11. MC validation • Future supercolliders will need predictive simulations • Good parametrisations can only be derived from good simulations! • To have some hope that a MC is predictive two kind of tests have to be performed exhaustively • Single process tests • Test beam validations • Experimental errors have to be considered attentively • Do not confuse systematic and statistical errors! • Very hard work… Innovative Detector for Supercolliders

  12. An example with GEANT4 • Low momentum hadrons are important for ALICE • Open geometry (no calorimeter to absorb particles) • Small magnetic field (0.4 T) • Account for most of the energy deposit • Particles "leaking" through the front absorbers and beam-shield generate background which limits the performance in central Pb-Pb collisions • In the forward direction also the high-energy hadronic collisions are of importance Innovative Detector for Supercolliders

  13. Proton Thin TargetExperimental Set-Up Data from Los Alamos in: Nucl. Sci. Eng., Vol. 102, 110, 112 & 115 Innovative Detector for Supercolliders

  14. Now Before Parameterised model:jpions: (p,Al) @ 597 MeV Innovative Detector for Supercolliders

  15. Double differentials Innovative Detector for Supercolliders

  16. Low energy neutrons Tiara facility Innovative Detector for Supercolliders

  17. Preliminary Results: 43 MeVTest Shield: Iron – Thickness: 20 cm Innovative Detector for Supercolliders

  18. Simulation and LCG • The LCG project is taking very seriously simulation activity • (At last!) a common project includes FLUKA and GEANT4 • A principle agreement has been found for the distribution of the FLUKA source • Single process and test beam validations are programmed for FLUKA and GEANT4 • Should have been started 10 years ago! • The project has decided to • Adopt the ALICE Virtual MonteCarlo as generic framework • Adopt the ALICE geometrical modeller Innovative Detector for Supercolliders

  19. But where to find the necessary computing power? • The needs of simulation are very large • For the next Data Challenge ALICE will need 1400kSI2K and 300TB for six months • Approximately 1400 high-end PCs running continuously! • But the computing needs to collect, store, reconstruct and analyse LHC data are even larger! Innovative Detector for Supercolliders

  20. The LHC Detectors CMS ATLAS ~6-8 PetaBytes / year ~109 events/year ~103 batch and interactive users LHCb Innovative Detector for Supercolliders

  21. CERN’s Network in the World Europe: 267 institutes, 4942 usersElsewhere: 208 institutes, 1752 users Innovative Detector for Supercolliders

  22. Why distributed computing? • The investment for LHC computing is massive • For ALICE only • 1.25 GB/s in HI mode • ~3 PB/y of tape • ~1.5 PB of disk • ~16,000 kSI2k (~16,000 PC2003) • ~ 8M€ of hardware • Without personnel + infrastructure and networking • Millions lines of code to develop and maintain for 20 years • Politically, technically and sociologically it cannot be concentrated in a single location • Whenever possible countries prefer national investments • Competence is naturally distributed • A concentrated facility would force people to travel to CERN often Innovative Detector for Supercolliders

  23. The distributed computing model • Every physicist should have equal access to data and resources • Resources for HEP computing (CPUs’ and data) will be distributed • Co-located in so-called regional centres • The centres should work as an integrated system to provide • Maximisation of the usage of the resources • Redundancy and fault tolerance • Security • Maximum transparency of usage • The system will be extremely complex • Number of sites & components in each site • Different tasks performed in parallel: simulation, reconstruction, scheduled and unscheduled analysis • Physicists have realised the challenge of this since few years • Studies started already some years ago (MONARC project) Innovative Detector for Supercolliders

  24. Lyon/ALICE RAL/ALICE OSC Tier3 University 1 Tier 1 centre 200 kSI95 300 TB disk Robot e.g. Catania 622 MB/s Tier 2 centre 20 kSI95 20 TB disk Robot Tier3 University 2 622 MB/s 1500 MB/s 622 MB/s e.g. Houston Univ Tier 2 centre 20 kSI95 20 TB disk Robot 622 MB/s CERN/ALICE Tier 0+1 centre 800 kSI95 500 TB disk Robot Tier3 University N The Monarc Model Innovative Detector for Supercolliders

  25. The challenge • Bad news is that the basic tools are missing (at production quality) • Distributed resource management • Distributed namespace for files and objects • Distributed authentication • Local resource management of large clusters • Data replication and caching • WAN/LAN monitoring and logging • Good news is that we are not alone • All the above issues are central to the new developments going on in the US and in Europe under the collective name of GRID Innovative Detector for Supercolliders

  26. The Grid:Blueprint for a New Computing InfrastructureI. Foster, C. Kesselman (Eds), Morgan Kaufmann, 1999 • Available July 1998; • ISBN 1-55860-475-8 • 22 chapters by expert authors including Andrew Chien, Jack Dongarra, Tom DeFanti, Andrew Grimshaw, Roch Guerin, Ken Kennedy, Paul Messina, Cliff Neuman, Jon Postel, Larry Smarr, Rick Stevens, and many others “A source book for the history of the future” -- Vint Cerf http://www.mkp.com/grids Innovative Detector for Supercolliders

  27. A Brief History • Early 90s • Gigabit testbeds, metacomputing • Mid to late 90s • Early experiments (e.g., I-WAY), academic software projects (e.g., Globus, Legion), application experiments • 2001 – now • Major application communities emerging • Major infrastructure deployments • Growing technology base • Global Grid Forum: ~500 people, >90 orgs, 20 countries • The “Grid problem” is about resource sharing & coordinated problem solving in dynamic, multi-institutional virtual organizations • Data is often the focus as opposed to classical numerically intensive simulations • Analogy with the power grid Innovative Detector for Supercolliders

  28. The GRID: networked data processing centres and ”middleware” software as the “glue” of resources. Researchers perform their activities regardless geographical location, interact with colleagues, share and access data Scientific instruments and experiments provide huge amount of data The Grid Vision Innovative Detector for Supercolliders

  29. A p p l i c a t i o n s Diverse global services Core Globus services Local OS GLOBUS hourglass, a model for middleware • Focus on architecture issues • Low participation cost, local control and support for adaptation • Use to construct high-level, domain-specific solutions • A set of toolkit services • Security (GSI) • Resource management (GRAM) • Information services (MDS) • Remote file management (GASS) • Communication (I/O, Nexus) • Process monitoring (HBM) Innovative Detector for Supercolliders

  30. Instruments are expensive Innovative Detector for Supercolliders

  31. Virtual Observatory: an exampleof the scientific opportunities • The planned Large Synoptic Survey Telescope will produce over 10PB/y by 2008! • All-sky survey every few days: fine-grain time series for the first time • The National Academy of Sciences recommends, as a first priority, the establishment of a National Virtual Observatory • http://www.nap.edu/books/0309070317/html/ Innovative Detector for Supercolliders

  32. Crab Nebula in 4 spectral regionsX-ray, optical, infrared, radio Innovative Detector for Supercolliders

  33. Virtual Sky: Image Federation http://virtualsky.org/ from Caltech CACR Caltech Astronomy Microsoft Research Virtual Sky has 140,000,000 tiles 140 Gbyte Change scale Change theme Optical (DPOSS) Xray (ROSAT) theme Innovative Detector for Supercolliders Coma cluster

  34. Not a centralized resource • The catalogs are NOT ingested into the Virtual Sky server, but left in place (maintained by their curators), and accessed by the accompanying web service • This is an important point: leave the resources and the responsibility for managing them where they would normally be Innovative Detector for Supercolliders

  35. Results Possible on TeraGrid Modelling Cell Structures • Pre-Blue Horizon: • Possible to model electrostatic forces of a structure with up to 50,000 atoms -- a single protein or small assembly • Pre-TeraGrid: • Possible to model one million atoms – enough to simulate drawing a drug molecule through a microtubule or tugging RNA into a ribosome • TeraGrid: • Models of 10 million atoms will make it possible to model function, structure movement, and interaction at the cellular level for drug design and to understand disease Baker, N., Sept, D., Joseph, S., Holst, M., and McCammon, J. A. PNAS98: 10037-10040 (2001). Innovative Detector for Supercolliders

  36. mammograms X-rays MRI cat scans endoscopies, ... Digital Radiology(Hollebeek, U. Pennsylvania) • Hospital Digital Data • Very large data sources - great clinical value to digital storage and manipulation and significant cost savings • 7 Terabytes per hospital per year • dominated by digital images • 2000 Hospitals x 7 TB per year x 2 = 28 PetaBytes per year Hierarchical Storage and Indexing Highly Distributed Source Innovative Detector for Supercolliders

  37. JDL/SandBox Output SandBox Resource brokering (easy) CE CE/SE CE/SE CE/SE CE/SE RB Run job CE/SE CE/SE CE Innovative Detector for Supercolliders

  38. Input “sandbox” DataSets info UI JDL Output “sandbox” grid-proxy-init SE & CE info Output “sandbox” Expanded JDL Job Submit Event Job Query Input “sandbox” + Broker Info Publish Job Status Storage Element Globus RSL Job Status Job Status Resource brokering (in reality) Replica Catalogue Information Service Resource Broker Author. &Authen. Job Submission Service Logging & Book-keeping Compute Element Innovative Detector for Supercolliders

  39. Grid projects • Several GRID projects have been launched by EU and US funding agencies • They have all started designing “the GRID” • Although based on common components such as GLOBUS • Tremendous richness of architectures and products • But worrying lack of stable testbeds where to experiment and provide feedback • At the moment only friendly and advanced users can use the system • Which of course creates a vicious circle… Innovative Detector for Supercolliders

  40. External software AliEn Core Components & services Interfaces RDBMS (MySQL) Database Proxy ADBI DBD DBI User Application File & Metadata Catalogue API (C/C++/perl) LDAP Authentication RB FS External Libraries User Interface Perl Modules Perl Core CLI Config Mgr CE SOAP/XML V.O. Packages & Commands SE GUI Web Portal Package Mgr (…) Logger Low level High level An alternative approach • Standards are now emerging for the basic building blocks of a GRID • There are millions lines of code in the OS domain dealing with these issues • Why not using these to build the minimal GRID that does the job? • Fast development of a prototype, no problem in exploring new roads, restarting from scratch etc etc • Hundreds of users and developers • Immediate adoption of emerging standards • An example, AliEn by ALICE (5% of code developed, 95% imported) Innovative Detector for Supercolliders

  41. OSU/OSC LBL/NERSC Dubna Houston Birmingham NIKHEF RAL Saclay Krakow GSI Nantes Karlsruhe Budapest CERN Merida Padova IRB Lyon Torino Bologna Bari Cagliari Yerevan Catania Kolkata, India Capetown, ZA AliEn activity http://www.to.infn.it/activities/experiments/alice-grid 32 sites configured 5 sites providing mass storage capability 12 production rounds 22773 jobs validated, 2428 failed (10%) Up to 450 concurrent jobs 0.5 operators Innovative Detector for Supercolliders

  42. Selection Parameters TagDB CPU Procedure PROOF DB1 RDB CPU Proc.C DB2 Proc.C DB3 CPU Proc.C DB4 CPU Proc.C DB5 CPU Proc.C DB6 CPU Distributed analysis model Local Remote Bring the KB to the PB and not the PB to the KB Innovative Detector for Supercolliders

  43. GriPhyN PPDG GRID iVDGL The GRID universe EGEE/LCG • The real problem is how to avoid divergence! Innovative Detector for Supercolliders

  44. MetaGrid and Grid federations • It has been realised that there will be many Grids and not a single one • However users will not want to learn more than one • The concept of Grid federation and Meta Grid are now explored • Unfortunately this sometimes looks like building on sand… • As we still do not have a stable base on which to build • And it does not help an early adoption and response from the users Innovative Detector for Supercolliders

  45. MetaGrid PseudoCE CE FedGrid1 PseudoCE PseudoCE CE CE CE CE AliEn stack Grid1 Grid2 CE FedGrid2 CE CE CE CE CE CE CE CE CE CE PseudoCE iVDGL stack EDG stack Meta and Federated grids AliEn User Interface Innovative Detector for Supercolliders

  46. OGSA A new standard emerging? OGSI Ansdm,nhnjkhdlasjdA KLAJSDLKJDKLAJDKLAJ JADSLJKALKK;ALSDK A;SLKDA;LDAS;LDK ASDK;LKASDL;ASDK;S AD;LKALSKDA;LDKSA SD;LKA;Lskda;lddlas Askd;laskdl;sakdl;askd Asdkl;askd;laskdl;askd Kas;ldka;lsdka;lskdk;ds • Grid Services are defined by OGSA: what Grid Services should be capable of, what types of technologies they should be based on • Grid Services are specified by OGSI, which is a formal and technical specification of the concepts described in OGSA, including Grid Services • Grid Services extend Web Services • Suppose you want to build a new house, you need • An architecture (OGSA) • A detailed engineering plan (OGSI) • Workers that build the hous (the real GRID) defines specifies Grid services Web services Standard interoperable technologies XML, WSDL, SOAP Innovative Detector for Supercolliders

  47. LCG new strategy • LHC experiments are proposing to LCG to • Start from the architecture of a demonstrated working product (strongly inspired from AliEn) • Develop components based on standards and/or adopt existing components • Deploy very quickly a prototype and expose it to the users • Refine iteratively the architecture and the services as needs and standards evolve • Don’t write doc, write working code! • Apparently LCG is receptive to these arguments Innovative Detector for Supercolliders

  48. Software development • In the LEP era the code was 90% written in FORTRAN • ~10 instructions, 50 pages! • In the LHC era the code is in many cooperating languages, mainly C++ • ~ 50 instructions, 700 pages – nobody understands it completely (B.Stroustrup) • Users are heterogeneous, sparse and without hierarchical structure • From very expert analysts to users, from 5% to 100% of time devoted to computing • People come and go with a very high rate • Programs have to be maintained by people who did not develop them • Young physicists need knowledge they can use also outside physics • Modern SE (“Agile Methodologies”) propose to value Individuals and interactions Working software Customer collaboration Responding to change processes and tools huge documentation contract negotiation following a plan OVER That is, while there is value in the items onthe right, we value the items on the left more. Innovative Detector for Supercolliders

  49. HEP, LHC & GRID • Funding for HEP is becoming scarce • There is a serious personnel deficit (also!) in HEP computing • The exceptional interest spreading in most countries for GRID resulted in an acute need for GRID-trained CSs • HEP (which invented the web) looked as an ideal place where to train young scientists in GRID technologies • This was seen as a unique opportunity to alleviate the personnel problems of LHC computing Innovative Detector for Supercolliders

  50. HEP, LHC & GRID • People are sent at CERN to work and train on GRID • LHC experiments may greatly profit from GRID • CERN has experience in large distributed collaborations • This could be a good deal, however • Mostly young and non-experienced CSs are sent at CERN • No knowledge of HEP habits and needs, little experience • CERN has no record in distributed computing research • LHC computing needs go beyond GRID middleware • There is a pressure to launder personnel into other roles • People come at CERN with agendas and constraints • GRID developed at CERN is specific for the need of HEP • However a working middleware for HEP would go a long way in satisfying most applications Innovative Detector for Supercolliders

More Related